Article 5.2 · The Collaboration Imperative: Why AI Needs Humans and Humans Need Each Other

T3PS Legal Dynamics · Series 5: The Human Factor · Refreshed June 2026

Let us start with an honest confession. I have watched the same dysfunction play out in litigation cost budgeting for twelve years — and I have watched it with increasing bewilderment, because it is so unnecessary and so persistent.

Two sets of solicitors, representing parties in a commercial dispute, are required to produce cost budgets and exchange them, discuss them, and where possible agree them. The Precedent H form has existed since 2013. The process is well-established. The purpose is clear: to ensure that litigation costs are proportionate and that both parties understand what they are committing to. Simple, in concept.

In practice, what happens is something closer to a financial arms race conducted entirely through PDFs.

The PDF Wars

The claimant sends a budget as a PDF. The defendant's solicitors annotate it, disagree with approximately 40% of the figures, and send back their objections — in a PDF. The claimant's solicitors respond with their own comments — in an email, with a revised PDF attached. The defendant's solicitors call this unacceptable and produce a counter-budget — in a PDF. By the time the parties attend the costs management conference, they have exchanged seventeen emails and nine PDFs, and the judge spends the first hour of a listed half-day hearing establishing what anyone actually agrees on.

Why are they not in a shared cloud document? Why has no one picked up the telephone and said: "Look, can we get on a call, share a screen, and work through this together? We both know roughly what this litigation should cost. Let's agree it and move on." There is no good reason. There are many bad ones: institutional habits, defensive professional behaviours, an adversarial instinct applied to a collaborative task, and — frankly — a billing incentive that rewards the hours spent on the process rather than the outcome the process is meant to produce.

All the AI in the world is not going to suddenly turn your law firm into the lean, mean billing machine you tell the golf club members it is. Not if the fundamental way you relate to other participants in the process remains unchanged.

The Real Collaboration Deficit

This matters beyond the specific frustrations of costs management. The legal profession has a collaboration deficit that predates AI, that AI investment will not automatically fix, and that needs to be honestly confronted.

Collaboration requires trust, which requires vulnerability, which requires the willingness to show what you know and what you do not know. The professional culture that treats information as competitive advantage — that keeps its cards face-down until forced to reveal them — is precisely the culture that makes AI adoption slow and shallow even in the firms that invest most heavily in the tools.

Consider how AI actually gets used effectively. It requires thought. It requires an understanding of what you are trying to produce, a clear brief, a process for checking the output against reality, a willingness to iterate when the first result is wrong. It requires the professional to engage with the tool as a thinking partner rather than a vending machine. That kind of engagement is, at its core, a collaborative skill — working with something else, being responsive to what it produces, being honest about when you need to push back.

The lawyer who is unwilling to collaborate with a colleague in the next office is unlikely to collaborate effectively with an AI tool. The reflexive habits of information hoarding, defensive communication and risk-averse non-engagement will simply be transferred to the new medium. They will use the AI to generate outputs quickly, but they will not do the deeper work of genuinely engaging with what it produces and asking whether it is right.

The Client Who Had Not Thought It Through

Here is a conversation I have had, in various forms, more than once. A client approaches, keen to adopt AI tools for their legal operation. The energy is high, the budget is allocated, the choice of platform has been made with considerable boardroom theatre.

"What do you want to use it for?" comes the question.

"Well — everything," comes the answer. "Documents, research, contracts, compliance."

"Which specific workflows are you looking to improve, and what does your current process look like?"

A pause. "I haven't really got time to work that out. I need it implemented."

This is not an unusual conversation. And it illustrates the fundamental misunderstanding that is driving a significant proportion of legal AI investment. AI is not a solution that you apply to problems. It is a capability that you direct towards problems that you have already understood. Buying the shiniest tool without a clear purpose is not transformation. It is expensive decoration.

A record £188.8 million was invested in UK legal tech in 2025. And yet, as one prominent talent partner noted following the LawtechUK data, only 24% of those investing firms had a formal plan for how to use their investment. The investment was landing. The transformation was not. Because transformation requires the harder work: process mapping, workflow redesign, change management, collaboration across teams that have historically operated in parallel silos.

AI amplifies existing capability. A firm with excellent processes, a collaborative culture and high-quality people will find AI multiplicative. A firm with poor processes, a siloed culture and inconsistent quality will find AI amplifies exactly those problems — faster document production of inconsistent quality, quicker research that nobody checks, automated workflows that embed the bad habits of the manual process.

Von Clausewitz and the Limits of Complex Plans

There is a passage in Clausewitz that has remained relevant for two centuries, and which applies to AI adoption in professional services with uncomfortable precision. In On War, Clausewitz writes: "The world has a way of undermining complex plans — particularly in fast-moving environments."

The legal technology landscape is moving fast. The firms that are trying to implement five-year AI transformation roadmaps developed in 2024 are already finding that the roadmap predates several significant developments. Harvey AI, which achieved an $11 billion valuation and over 400,000 agentic queries daily, was not the product described in most firms' AI strategies when those strategies were written. Agentic AI — systems that can take sequences of actions autonomously rather than responding to single prompts — has changed the capability landscape faster than most technology committees anticipated.

The response to this is not to give up on planning but to build for adaptability rather than certainty. The firms that are navigating this well are not the ones with the most detailed roadmaps. They are the ones with cultures that can respond quickly to new information, make decisions without requiring three committees to sign off, and actually talk to each other about what is working and what is not. Which brings us back, as it always does, to collaboration.

The hybrid model — AI capability combined with human creativity, empathy and ethical judgment — is not a compromise position between the technologists and the traditionalists. It is the only model that actually works. AI without human judgment produces confident errors. Human judgment without AI-driven capability is, in the current environment, simply slower and more expensive than it needs to be. The question is not which one wins. The question is how the combination is managed, and who is responsible for making sure it is managed well.

The Real Competition

The framing that the legal profession needs to let go of is the one that treats AI as a threat to individual lawyers and law firms from outside the profession. The real competitive dynamic is entirely internal: who can leverage AI most effectively to provide better services, not who can bill the most hours.

Research consistently shows that AI has enhanced rather than replaced client-professional relationships when it is deployed thoughtfully. The lawyer who uses AI to do the routine work faster has more time for the client conversations that matter. The client who receives faster, clearer, better-organised work product trusts the lawyer more. The relationship deepens rather than becoming more transactional.

But this only happens if the human element is genuine. If the lawyer uses the time freed by AI to generate more billable work rather than to deepen the client relationship, nothing changes strategically. If the firm treats AI adoption as a cost reduction exercise rather than a quality enhancement exercise, the client eventually notices. The efficiency gains are real; whether they translate into better client outcomes or simply into higher partner profits is a choice — and it is a choice that the profession has a poor track record of making generously.

Collaboration as Competitive Strategy

The adversarial instinct that runs deep in legal culture is appropriate in some contexts and deeply counterproductive in others. In a trial, adversarial is the point. In costs management, in disclosure protocols, in agreed orders, in anything that both parties ultimately want resolved efficiently — adversarial is waste. It costs everyone, including the client who pays for the conflict.

The question worth asking, as AI enters the profession with genuine force, is whether the tools will be used to entrench those adversarial habits or to challenge them. There is a version of AI adoption in litigation that makes the PDF wars faster and more elaborate. There is another version — shared platforms, transparent data, collaborative budgeting tools that both sides can work on simultaneously — that makes them unnecessary.

The humans have to choose which version they want. The algorithm will serve whichever purpose it is given.

Twelve years since CPR Part 3 and the Precedent H regime created the costs management framework, and we are still sending each other PDFs. That is not a technology failure. That is a collaboration failure. And no AI platform, however sophisticated, will fix it until the humans on both sides of the matter decide that the purpose of costs management is to reach an agreed, proportionate budget — not to win the preliminary skirmish.

When that decision is made — when the phone call replaces the email chain, when the shared document replaces the amended PDF, when the budget discussion is a conversation rather than a correspondence — AI can make the underlying work dramatically better. Better data, better forecasting, better modelling of different litigation scenarios. The tools are genuinely capable. The humans need to decide they want to use them in that spirit.

The collaboration imperative is not about being nice. It is about being effective. And the firms that understand this — that treat AI as a tool for achieving better outcomes through better collaboration, rather than as a mechanism for doing the same things faster in isolation — are the ones that will look back on this period of transformation with something approaching satisfaction.

The rest will be spending the year 2030 exchanging PDFs and wondering why their clients have started going elsewhere.

Questions worth sitting with:

1.    If costs management under Precedent H requires genuine collaboration to work properly, what would it take — procedurally, culturally, perhaps even by rule — to shift the default from adversarial to cooperative?

2.    When your firm says it is "implementing AI," what does it mean in terms of the specific workflows that have actually changed — and how would you measure whether those changes have improved outcomes for clients rather than simply reduced costs for the firm?

3.    If AI amplifies existing capability, what does that imply about the most important investment a law firm should be making right now — and is it a technology investment at all?